[TIP] mock.patch in version 0.5 doesn't mock __exit__ method on open files

[Michael Foord]

Hello Matthew,

Technically this is a bug in Python (which will be fixed in 2.7 / 3.2) 
where __enter__ / __exit__ are looked up on the instance rather than the 
class.

That doesn't help you however. :-)

Not sure why this works in 0.4 and fails in 0.5 - it shouldn't have 
worked in 0.4 either. Mocking magic methods (awesome alliteration) is a 
tricky one as it will need custom support in Mock. In general I don't 
think the magic methods should exist on mock objects unless the 
programmer has configured them, so I need a simple way for the user of 
mock to specify which magic methods should exist.

Mock is due for an update so I will think about this. I'll work out a 
suggested API for your feedback. I guess you want __enter__ to return a 
mock object that you can track and __exit__ to record calls to and 
return a boolean.

All the best,

Michael

Matthew Wilson wrote:
> This test passes when I use mock 0.4 and fails with 0.5:
>
> @patch('__builtin__.open')
> def test_to_html(o):
>
>     global b
>     b.to_html('bogus filepath')
>
> Here's the definition of b.to_html:
>
>     def to_html(self, filepath):
>         """
>         Write this bag out as HTML to a file at filepath.
>         """
>
>         with open(filepath, 'w') as f:
>             f.write(self.html)
>
> mock 0.5 has this code:
>
>     128     def __getattr__(self, name):
>     129         if self._methods is not None:
>     130             if name not in self._methods:
>     131                 raise AttributeError("Mock object has no
> attribute '%s'" % name)
>     132         elif _is_magic(name):
> --> 133             raise AttributeError(name)
>
> When __exit__ gets called, that code raises the AttributError.
>
> I can see good reasons fo not mocking magic methods automatically, but
> can I explicitly say I do want __open__ and __exit__ to be mocked?
>
> mock is great software by the way.
>
> Matt
>
>
>   


-- 
http://www.ironpythoninaction.com/
http://www.voidspace.org.uk/blog